The Agent Skills Directory

[SAFE]: The skill implements a robust documentation lifecycle management workflow. It performs standard file and git operations restricted to the project repository's scope.
[COMMAND_EXECUTION]: The skill utilizes dedicated search and read tools (Glob, Grep, Read) for investigation. It explicitly instructs subagents to avoid shell commands like ls, find, or bash for file operations, which reduces the risk of unintended command execution or shell injection.
[DATA_EXFILTRATION]: No data exfiltration vectors were detected. File access is limited to the docs/solutions/ directory and memory files. Network activity is limited to standard git operations for committing changes and opening pull requests.
[PROMPT_INJECTION]: No patterns of malicious prompt injection or behavior overrides were found. The instructions are focused on the logic of document classification (Keep, Update, Consolidate, Replace, Delete) and user interaction modes.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-contributed documentation files that could theoretically contain instructions, the risk is mitigated by the highly structured processing logic and the use of isolated subagents for investigation and replacement tasks.

ce-compound-refresh